Analysis of Variance (ANOVA)
Randomized Block Design
First, let’s consider the assumptions
(Handouts: Assumptions Handout)
When using one-way analysis of variance, the process of
looking up the resulting value of F in an F-distribution
table, is reliable under the following assumptions:
• The values in each of the groups (as a whole) follow the
normal curve,
• with possibly different population averages (though the
null hypothesis is that all of the group averages are
equal) and
• equal population variances.
Normal Distribution
• While ANOVA is fairly robust with respect to
normality, a highly skewed distribution may have
an impact on the validity of the inferences
derived from ANOVA.
• Check this with a histogram, stem-and-leaf
display, or normal probability plot for the
response, y, corresponding to each treatment.
Equal Population Variances
• Book lists a few formal tests of homogeneity of
variances available (p. 635, 636)
• We can approximately check this by using as
bootstrap estimates the sample standard
deviations.
• In practice, statisticians feel safe in using ANOVA
if the largest sample SD is not larger than twice
the smallest.
Randomized (Complete) Block Design
• Sample Layout: Each horizontal row represents a
block. There are 4 blocks (I-IV) and 4 treatments (A-D)
in this example.
Block I
Block II
Block III
Block IV
ABCD
DABC
BDCA
CABD
Randomized Block Designs (Formulas: p. 575)
Total SS = Treatment SS + Block SS + Error SS
SS(Total) = SST + SSB + SSE
In Minitab, we will use the Two-Way ANOVA option or
the General Linear Model option.
We will have two F values
• For testing treatments:
MST
F
MSE
MSB
• For testing blocks: F 
MSE
These values are equivalent to the values in our
nested F
Why test blocks?
• If we want to determine if blocking was effective
in removing extraneous sources of variation
– Is there evidence of difference among block means?
• If there are no differences between block means,
we lose information by blocking
– Blocking reduces the number of degrees of freedom
associated with the estimated variance of the model
• BUT it is okay to use the block design if you
believe they may have an effect (p. 574)